Driver Fatigue Driving Monitoring System Based On The Computer Vision

Fatigue driving has become the biggest danger in traffic accidents. Frequent traffic accident surrounds us worldwide and high number of malignant accidents has been occurred. Study shows that most of them are due to the driver's tiredness. So the driver safety assist systems specific to driver fatigue driving has become research hot topic at home and abroad. This paper designs a safety monitoring system base on computer vision, in order to prevent the condition of fatigue driving.The paper select PERCLOS which is the best evaluation method as the judgement method. It use eyes closing to a certain degree and achieve at a certain length of time to determine the condition, and this paper select P70 as the judgement standard, namely that eyes close to 70 percent.The first problem to solve is face detection, when using computer vision to design tiredness driving monitoring system. This paper process illuminate compensation in YCbCr color space first, and receive an even light distribution color image. Then use gaussian model as the skin model to separate face image from the driver image, and gain its aim that separate face image from background. There will be some disruptive signals that similar to the colour of skin. This paper designs an object centroid localization method to remove the interference information from face image, and receive a good detection effect.Another major problem is the issue of human eye's orientation. This paper designs an integral projection method, that use gray image to do integral projection along the horizontal and vertical direction. Because the gray value of eye position are small and other pixel gray value of face image are bigger, eye position is where the curved line has big inequality. Then use mathematical method to separate the eye position from the face image accurately. And use Canny operator to get the edge image of the eye image. At this time, use P70 of PERCLOS method to judge the driver's condition. The fatigue monitoring system designed this paper can detect driver's state in the unosculatory condition, and test results are accurate and timely.